Running 10 Projects at Once with AI — Not Easier, Just More Bearable

"You develop with AI? So how many things are you running at once?"

I get this question a lot these days. The honest answer: right now I'm running somewhere between 6 and 10 projects and client engagements in parallel. But there's one thing people almost always get wrong when I say that number — they assume AI made it easy.

It didn't. It made it more bearable. That distinction is the whole point of this post.

How I actually run things in parallel

First, the mechanics. I spin up an isolated workspace (a git worktree) per project and run multiple sessions in parallel. Whatever correction or judgment call I learn in one project gets written down as memory, so the next session doesn't repeat the same mistake. And every code change has to pass a type-check-and-test gate before a turn is considered done.

Because of this structure, switching from project A to project B doesn't mean re-explaining context from scratch, and there's always a minimum safety net catching mistakes. At this point it's tempting to conclude "so it really is easy now."

But what stays hard, stays hard

What a parallel system like this removes is repeated work and the cost of re-explaining context. What it does not remove:

• The sheer weight of having that many things open at once. Each project got lighter individually, but the feeling of 6-10 things simultaneously sitting in your head "waiting on a decision" doesn't go away.
• Judgment is still on me. Whether a requirement is actually the right call, whether a piece of code is really ready to ship — AI doesn't take that responsibility off my hands. More projects just means more moments that need judgment.
• Switching cost went down, not away. Moving my head from project A's context to project B's context still takes time and energy. The tool holds the notes for me; it doesn't do the mental switch for me.

So whenever I hear "you run 10 things at once with AI," I mentally rephrase it: "AI made 10 things survivable to run."

Working alone hurts regardless of the count

Working alone is hard even with a single project, because being the only person making the call also means being the only person catching a bad call. Scale that to 6-10 projects and the pain doesn't disappear — it multiplies in parallel. The tools cut down repetition and mistakes inside each project, but they don't change the underlying structure of carrying everything alone across projects.

Which is why you eventually need a team

My conclusion is simple. AI increasing how much one person can do is not a reason to keep doing it alone — if anything, it's the opposite. Once the tools shrink the repetition cost, what's left is the pure volume of judgment and responsibility, and that's the part that only gets lighter when more people share it.

"I can do 10 things alone" doesn't mean "I can do 10 things easily." It's closer to "I can grind through 10 things hard enough to survive." Whether you keep grinding that endurance or bring in more people to split the weight — that's the real choice a solo developer faces in the AI era.

AI made multiplaying possible. Whether you keep carrying that multiplay alone is a separate question.